Air conditioner defrosting method based on AI image processing

Abstract

The invention discloses an AI image processing-based air conditioner defrosting method, which comprises the following steps that a bracket and an image recognition module are mounted below a top cover of an air conditioner host, and the image recognition module is arranged on the bracket; the image recognition module obtains a real-time image of the fin coil pipe in the working state of the air conditioner according to a preset time interval or instruction; comparing the real-time image with a reference image of the finned coil to obtain distinguishing features; and according to the distinguishing characteristics, a defrosting instruction or no instruction is sent to a main control system of the air conditioner. The problems of frost-free defrosting, long-time non-defrosting, inaccurate defrosting starting, inaccurate defrosting condition judgment and incapability of accurately judging whether defrosting is completed or not are solved, energy is effectively saved, consumption is effectively reduced, the operation efficiency of the air conditioner is improved, and the service life of the air conditioner is prolonged.

Description

Technical Field

[0001] The present invention relates to the technical field of air-conditioning intelligent control, and in particular to an air-conditioning defrosting method based on AI image processing. Background Art

[0002] There are two main methods for defrosting control of existing air source heat pump air conditioning units: one is a control method based on time parameters, and the other is an intelligent control method based on parameters such as temperature, humidity and pressure.

[0003] When the air source heat pump air conditioning unit is in heating condition, the moisture in the air will adhere to the surface of the fin coil when it evaporates and absorbs heat to form a frost layer, which greatly reduces the heat exchange efficiency of the fin coil. It will further form an ice layer and make the heat pump air conditioner ineffective. Therefore, the defrost procedure must be performed to ensure the normal operation of the air conditioning heat pump; the defrost start time and duration vary irregularly with the temperature and humidity of the outdoor environment.

[0004] The time parameter defrost method has simple logic, is easy to implement and has low implementation cost. Most air conditioner outdoor units use this method for defrosting, but the disadvantage is obvious: defrost is often turned on when there is no frost, and defrost is not turned on when there is frost or even ice.

[0005] The intelligent defrosting method based on temperature, humidity and pressure parameters can determine the defrosting start and stop conditions according to the changes in parameters such as ambient temperature and humidity and the temperature and pressure of the refrigerant. It can achieve intelligent defrosting to a certain extent, but the control logic is complex and the implementation is slightly difficult. It is also prone to misjudgment and cannot defrost normally due to the complex changes in ambient temperature and humidity and the changes in the operating load of the air-conditioning host. There is also a method of detecting the air pressure difference on both sides of the fin coil to assist in solving the problem of energy waste and lack of comfort caused by inaccurate judgment of frosting conditions, but it can only solve the problem of inaccurate judgment under specific operating conditions of fixed-frequency air-conditioning units, but it is helpless for the increasingly widely used variable-frequency air source heat pump air conditioners. Summary of the invention

[0006] According to an embodiment of the present invention, a method for defrosting an air conditioner based on AI image processing is provided, comprising the following steps:

[0007] An image recognition module and a bracket are installed under the top cover of the air conditioner main unit, and the image recognition module is set on the bracket;

[0008] The image recognition module obtains a real-time image of the fin coil in the working state of the air conditioner according to a preset time interval or instruction;

[0009] Compare the real-time image with a reference image of the fin coil to obtain distinguishing features;

[0010] According to the distinguishing characteristics, a defrost command is issued to the main control system of the air conditioner or no command is issued.

[0011] Furthermore, the image recognition module includes a smart camera and an edge computing chip, the smart camera is used to obtain real-time images of the fin coil, and the edge computing chip is used to obtain distinguishing features and send a defrost command to the main control system of the air conditioner or not send any command.

[0012] Furthermore, the image recognition module includes an intelligent camera and a data transmission module. The intelligent camera is used to obtain real-time images of the fin coil. The data transmission module transmits the real-time images to an external operation center. The operation center obtains distinguishing features and sends a defrost command to the main control system of the air conditioner or does not send any command.

[0013] Further, obtaining the distinguishing features includes the following steps:

[0014] Acquire a frost-free fin coil image as a reference image;

[0015] The image characteristics of the reference image and the real-time image are used as parameter extraction objects;

[0016] Convert image characteristics into benchmark parameters that can be quantified and compared, extract image characteristics of real-time images and parameterize them into quantitative parameters;

[0017] The quantitative parameter is compared with the reference parameter to obtain the parameter difference, which is the distinguishing feature.

[0018] Furthermore, image characteristics include but are not limited to: edge information, texture changes, and color distribution.

[0019] Furthermore, extracting image features includes the following steps:

[0020] Preprocessing the reference image;

[0021] The edge detection algorithm is used to extract the edge intensity information in the preprocessed reference image to identify the boundary of the frost area;

[0022] Calculate the gray-level co-occurrence matrix of the image to determine the texture changes in the frost and ice area;

[0023] If the reference image is a color image, the color distribution information is extracted to distinguish the differences between different materials;

[0024] Construct and train convolutional neural network models on edge information, texture changes, and color distribution to learn and extract image features.

[0025] Furthermore, the image characteristics are converted into benchmark parameters that can be quantified and compared, and the quantified parameters are obtained, which includes the following steps:

[0026] Obtaining the proportion of the frost area to the area of ​​the reference image according to edge information, texture change and color distribution, or, according to edge information, texture change and color distribution, presetting the impact score of the frost degree to obtain the reference parameter;

[0027] Establish and calibrate a standards database on benchmark parameters;

[0028] A prediction model is built based on a standard database, taking image features as input and outputting quantitative parameters.

[0029] Furthermore, the prediction model is divided into a training set and a test set for cross-validation and optimization of the prediction model.

[0030] Furthermore, the reference image is preprocessed by gray-scaling, denoising and enhancing operations, the denoising operation is performed by a median filter or a Gaussian filter, and the enhancement operation is performed by using histogram equalization.

[0031] Further, whether to send a defrost instruction to the main control system of the air conditioner includes the following steps:

[0032] Preset standard parameter difference;

[0033] If the parameter difference exceeds the standard parameter difference, a defrost command is sent to the air conditioning main control system, otherwise, no command is sent.

[0034] The air conditioner defrosting method based on AI image processing according to the embodiment of the present invention solves the problems of defrosting without frost, no defrosting for a long time, inaccurate defrosting start, inaccurate judgment of defrosting conditions and inability to accurately judge whether defrosting is completed, effectively saves energy and reduces consumption, and improves the operating efficiency and service life of the air conditioner.

[0035] It is to be understood that both the foregoing general description and the following detailed description are exemplary, and are intended to provide further explanation of the technology as claimed. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a flow chart of an air conditioner defrosting method based on AI image processing according to an embodiment of the present invention;

[0037] Figure 2 for Figure 1 Flowchart of sub-steps of S3;

[0038] Figure 3 for Figure 2 The sub-step flow chart of S32;

[0039] Figure 4 for Figure 2 The sub-step flow chart of S33;

[0040] Figure 5 for Figure 1 Flowchart of the sub-steps of S4. DETAILED DESCRIPTION

[0041] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings to further illustrate the present invention.

[0042] First, combine Figures 1 to 5 An air conditioner defrosting method based on AI image processing according to an embodiment of the present invention is described, which is used for precise defrosting of an air conditioner and has a wide range of application scenarios.

[0043] like Figure 1 As shown, the air conditioner defrosting method based on AI image processing in an embodiment of the present invention has the following steps:

[0044] Specifically, Figure 1 As shown, in S1, an image recognition module and a bracket are installed under the top cover of the air-conditioning main unit. The image recognition module is set on the bracket. The bracket can be installed at any position under the top cover of the air-conditioning main unit, as long as it is easy to install and easy to take pictures of the fin surface. The application is simple and easy to install and maintain.

[0045] In this embodiment, the image recognition module can adopt two layout schemes: central computing and edge computing.

[0046] Central computing solution:

[0047] The image recognition module includes an intelligent camera and a data transmission module. The intelligent camera is used to obtain real-time images of the fin coil. The data transmission module transmits the real-time images to an external operation center. The operation center obtains distinguishing features and sends a defrost command to the main control system of the air conditioner or does not send any command.

[0048] Edge computing solutions:

[0049] The image recognition module includes a smart camera and an edge computing chip. The smart camera is used to obtain real-time images of the fin coil, and the edge computing chip is used to obtain distinguishing features and send a defrost command to the main control system of the air conditioner or not send any command.

[0050] Regardless of whether a central or edge computing solution is used, the smart camera can also be replaced with other imaging devices as needed, as long as the image quality is high enough to clearly show the frost conditions.

[0051] Specifically, Figure 1 As shown, in S2, the image recognition module obtains a real-time image of the fin coil in the air-conditioning working state according to a preset time interval or instruction.

[0052] Specifically, Figure 1As shown, in S3, the real-time image is compared with the reference image of the fin coil to obtain distinguishing features.

[0053] Further, obtaining the distinguishing features includes the following sub-steps:

[0054] like Figure 2 As shown, in S31, a frost-free fin coil image is acquired as a reference image.

[0055] like Figure 2 As shown, in S32, the image characteristics of the reference image and the real-time image are used as parameter extraction objects. In this embodiment, the image characteristics include but are not limited to: edge information, texture change, and color distribution.

[0056] Further, if Figure 3 As shown, extracting image features includes the following sub-steps:

[0057] like Figure 3 As shown, in S321, the reference image is preprocessed; in this embodiment, the reference image is preprocessed by grayscale, denoising, and enhancing operations to improve the efficiency and accuracy of subsequent processing, and the denoising operation is performed by a median filter or a Gaussian filter, and the enhancement operation is performed by using a histogram equalization technique. Of course, in this embodiment, more processing means can be introduced as needed to further optimize the preprocessing effect.

[0058] like Figure 3 As shown, in S322, an edge detection algorithm is used. In this embodiment, the Sobel operator and the Canny algorithm are used to extract edge intensity information in the preprocessed reference image to identify the boundary of the frost area.

[0059] like Figure 3 As shown, in S323, the gray level co-occurrence matrix (GLCM) of the image is calculated to determine the texture change of the frost area.

[0060] like Figure 3 As shown, in S324, if the reference image is a color image, color distribution information is extracted to distinguish the differences between different materials (such as metal and frost and ice).

[0061] like Figure 3 As shown, in S325, a convolutional neural network model about edge information, texture changes and color distribution is constructed and trained to learn and extract image characteristics.

[0062] like Figure 2 As shown, in S33, the image characteristics are converted into reference parameters that can be quantified and compared, and the image characteristics of the real-time image are extracted and parameterized into quantitative parameters. The specific sub-steps are as follows:

[0063] like Figure 4 As shown, in S331, the impact score of the frost degree is preset according to the edge information, texture change and color distribution as a comprehensive scoring system, or the proportion of the frost area to the area of ​​the benchmark image is obtained according to the edge information, texture change and color distribution to obtain the benchmark parameters.

[0064] like Figure 4 As shown, in S332, a standard database for benchmark parameters is established and calibrated, and images under various known frost conditions and their corresponding quantitative parameters are collected for model training and verification, and to ensure that measurement results under different conditions are comparable.

[0065] like Figure 4 As shown, in S333, a prediction model is constructed based on a standard database, image characteristics are used as input, and quantitative parameters are output. In this embodiment, the prediction model is divided into a training set and a test set, so as to achieve cross-validation and optimize the prediction model.

[0066] like Figure 2 As shown, in S34, the quantization parameter is compared with the reference parameter to obtain a parameter difference, and the parameter difference is the distinguishing feature.

[0067] Specifically, in S4, according to the distinguishing features, a defrost command is issued to the main control system of the air conditioner or no command is issued, which specifically includes the following sub-steps:

[0068] like Figure 5 As shown, in S41, a standard parameter difference is preset.

[0069] like Figure 5 As shown, in S42, if the parameter difference exceeds the standard parameter difference, a defrost command is sent to the air-conditioning main control system, otherwise, no command is sent.

[0070] Therefore, the trained model is deployed in the actual application scenario to monitor the frost and ice conditions of the air-conditioning outdoor unit in real time. The air-conditioning operation mode can be automatically adjusted according to the monitoring results, such as starting the defrost program, etc., which solves the problems of frost-free defrosting, no defrosting for a long time, inaccurate defrost start, inaccurate defrost condition judgment, and inaccurate judgment of whether defrosting is completed. It effectively saves energy and reduces consumption, and improves the operating efficiency and service life of the air conditioner.

[0071] Above, refer to Figures 1 to 5 An air conditioner defrosting method based on AI image processing according to an embodiment of the present invention is described, which solves the problems of defrosting without frost, no defrosting for a long time, inaccurate defrosting start, inaccurate defrosting condition judgment and inability to accurately judge whether defrosting is completed, effectively saves energy and reduces consumption, and improves the operating efficiency and service life of the air conditioner.

[0072] It should be noted that, in this specification, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprises..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0073] Although the content of the present invention has been described in detail through the above preferred embodiments, it should be appreciated that the above description should not be considered as a limitation of the present invention. After reading the above content, it will be apparent to those skilled in the art that various modifications and substitutions of the present invention will occur. Therefore, the protection scope of the present invention should be limited by the appended claims.

Claims

1. An air conditioner defrosting method based on AI image processing, characterized in that: The following steps are included: An image recognition module and a bracket are installed below the top cover of the air conditioner main unit, wherein the image recognition module is arranged on the bracket; The image recognition module acquires a real-time image of the fin coil in the air conditioner working state according to a preset time interval or instruction; Comparing the real-time image with a reference image of the fin coil to obtain distinguishing features; According to the distinguishing features, a defrost command is issued to the main control system of the air conditioner or no command is issued.

2. The air conditioner defrosting method based on AI image processing as claimed in claim 1, characterized in that: The image recognition module includes a smart camera and an edge computing chip. The smart camera is used to obtain a real-time image of the fin coil, and the edge computing chip is used to obtain the distinguishing features and send a defrost command to the main control system of the air conditioner or not send any command.

3. The air conditioner defrosting method based on AI image processing as claimed in claim 1, characterized in that: The image recognition module includes an intelligent camera and a data transmission module. The intelligent camera is used to obtain a real-time image of the fin coil. The data transmission module transmits the real-time image to an external operation center. The operation center obtains the distinguishing features and sends a defrost command to the main control system of the air conditioner or does not send any command.

4. The air conditioner defrosting method based on AI image processing as claimed in claim 1, characterized in that: Obtaining distinguishing features includes the following steps: Acquire a frost-free fin coil image as a reference image; Taking the image characteristics of the reference image and the real-time image as parameter extraction objects; Convert the image characteristics into reference parameters that can be quantified and compared, extract the image characteristics of the real-time image and parameterize them into quantitative parameters; The quantization parameter is compared with the reference parameter to obtain a parameter difference, and the parameter difference is the distinguishing feature.

5. The air conditioner defrosting method based on AI image processing as claimed in claim 4, characterized in that: The image characteristics include but are not limited to: edge information, texture changes, and color distribution.

6. The air conditioner defrosting method based on AI image processing as claimed in claim 5, characterized in that: Extracting the image characteristics comprises the following steps: Preprocessing the reference image; Using an edge detection algorithm to extract edge intensity information from the preprocessed reference image to identify the boundary of the frost and ice area; Calculating a gray level co-occurrence matrix of the image to determine texture changes in frost and ice areas; If the reference image is a color image, extracting color distribution information to distinguish differences between different materials; Construct and train a convolutional neural network model about edge information, texture changes, and color distribution to learn and extract the image characteristics.

7. The air conditioner defrosting method based on AI image processing as claimed in claim 6, characterized in that: Converting the image characteristics into benchmark parameters that can be quantified and compared, and obtaining the quantified parameters, comprises the following steps: Obtaining the proportion of the frost area to the area of ​​the reference image according to the edge information, texture change and color distribution, or, according to the edge information, texture change and color distribution, presetting the impact score of the frost degree to obtain the reference parameter; Establishing and calibrating a standard database regarding said reference parameters; A prediction model is constructed based on the standard database, the image characteristics are taken as input, and the quantization parameters are output.

8. The air conditioner defrosting method based on AI image processing as claimed in claim 7, characterized in that: The prediction model is divided into a data set into a training set and a test set for cross-validation and optimization of the prediction model.

9. The air conditioner defrosting method based on AI image processing as claimed in claim 6, characterized in that: The reference image is preprocessed by grayscale conversion, denoising and enhancement operations, the denoising operation is performed by a median filter or a Gaussian filter, and the enhancement operation is performed by histogram equalization.

10. The air conditioner defrosting method based on AI image processing as claimed in claim 4, characterized in that: Whether to send a defrost command to the main control system of the air conditioner includes the following steps: Preset standard parameter difference; If the parameter difference exceeds the standard parameter difference, a defrost command is sent to the air-conditioning main control system; otherwise, no command is sent.

Images